Research On Frontier-based Robot Autonomous Exploration Approaches

Robot autonomous exploration could map the environment independently and autonomously,which provides the basis for robots' further localization and navigation.Thus it is a key step toward real robotic autonomy.Among various approaches for robot autonomous exploration,frontier-based methods are most commonly used,and the core of this kind of method is how to quickly find frontiers(the boundary between the known obstacle free area and the unknown area)in the current environmental map.One efficient method of frontier detection exploits the idea of the rapidly-exploring random tree and uses tree edges to search for frontiers.Compared to traditional frontier detection methods based on edge detection of image processing,it can be applied to search frontiers in largescale and high-dimensional environmental maps more efficiently.However,this method needs to occupy a large number of storage resources and searches frontiers slowly in the environment where the traditional rapidly-exploring random tree is not easy to grow(unfavorable environment).In this thesis,a sampling-based multi-tree fusion algorithm(SMF)for frontier detection is proposed,which is aiming at solving the problems of the above frontier detection methods.Our frontier detection method is also based on the idea of the rapidlyexploring random tree,but we fully realize the difference between functions of the rapidly-exploring random tree used in searching frontiers and in generating feasible robot motion paths.Thus we change the growing rule of the rapidly-exploring random tree.When newly grown nodes and edges fall in the known area,even if they cross obstacles,we still regard these tree nodes as valid and remain them.Because the growing process of modified RRT does not need to avoid obstacles in the known area,the problem that it grows slowly and searches frontiers slowly in an unfavorable environment can be solved.At the same time,because the frontier detection does not need to retain the feasible path information of the robot by retaining the edge information as in the path planning,this thesis will discard the edge information after using the new grown edge to search frontiers and only retain the information of tree nodes to grow new tree nodes later,thus saving memory resources.Secondly,because the external tree nodes near frontiers in the rapidlyexploring random tree play a more important role in frontier search than the internal tree nodes.This thesis proposes a block structure to delete a large number of redundant tree nodes in the internal blocks,thus greatly reducing the storage resource requirements of our algorithm.At last,two kinds of rapidly-exploring random trees with different growth modes are fused in this thesis,so as to accelerate the search speed of frontiers which are near to the robot or just created by new obtained environmental information on the basis of ensuring that all frontiers in the map can be found.In this thesis,many experiments are carried out in simulation environments and the real environment to verify that the proposed sampling-based multi-tree fusion algorithm for frontier detection not only saves the memory resources without losing the exploration performance,but also has a better performance in unfavorable environments where the traditional rapidly-exploring random tree is not easy to grow.